研究大模型在算术任务中的策略不稳定性，揭示其零样本能力受限的原因。
Algorithmic Phase Transitions in Language Models: A Mechanistic Case Study of Arithmetic
- 通过分析模型在不同算术任务中的解题策略变化，揭示算法不稳定性现象。
- Gemma-2-2b 在四位与八位加法任务中采用截然不同的计算方式。
- 适合关注模型可解释性与推理泛化能力的研究者阅读。
大型语言模型的零样本能力使其能在未显式训练的情况下解决多种任务，但其为何能完成某些任务而无法完成其他任务仍不明确。本文通过定义并研究语言模型中的算法稳定性——即任务设定变化导致模型问题求解策略的变化——来揭示这一现象。我们聚焦于需要算法稳定才能泛化的任务：双操作数算术。令人惊讶的是，Gemma-2-2b 在四数字与八数字加法任务中采用了显著不同的计算模型。结果表明，算法不稳定性可能是语言模型在特定逻辑推理任务上零样本表现不佳的一个因素，因其难以抽象不同求解策略并平滑过渡。
原文摘要 · Abstract (English)
Zero-shot capabilities of large language models make them powerful tools for solving a range of tasks without explicit training. It remains unclear, however, how these models achieve such performance, or why they can zero-shot some tasks but not others. In this paper, we shed some light on this phenomenon by defining and investigating algorithmic stability in language models -- changes in problem-solving strategy employed by the model as a result of changes in task specification. We focus on a task where algorithmic stability is needed for generalization: two-operand arithmetic. Surprisingly, we find that Gemma-2-2b employs substantially different computational models on closely related subtasks, i.e. four-digit versus eight-digit addition. Our findings suggest that algorithmic instability may be a contributing factor to language models' poor zero-shot performance across certain logical reasoning tasks, as they struggle to abstract different problem-solving strategies and smoothly transition between them.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。